The issue is it is not correct, and people being people, they are going to start trolling and calling people out when it might not be true. This is a slippery slope.
I have entered my "How I made this ... " in full snark mode, but people screaming about writers using AI by using an AI tool is just ... dumb ... in my opinion
As I've said elsewhere, my main concern is that it could falsely accuse writers of using AI, as these tools aren't reliable and often have false positives. However, one good thing is that people will probably be much less likely to use AI (and there will be a lot less slop).
HI Kristina. I’ve spouted off enough about this in the past 48 hours but your thoughtful post and the inputs from Patrick and others are helpful color.
I’ve no issue with Substack making this a feature. If an Author/Publisher FEELS they want to have this capability on their publications that’s fine. But it should not be a global feature. Making it a manual opt out EVER atone is a degree of friction that is either designed in deliberately OR the thoughtless product management.
They need to rectify this. and put it back in the hands of the creators decision.
On the contractual position and lazy SLA language and process to notify users this has become extremely prevalent in the new generation of companies. Old norms of what a customer relationship meant and looked like seem to be being deprecated daily. Sad 😞
Good questions, disturbing news about the failure to inform members of the contractual changes. I'm following with interest. I won't use Pangram to check other people's work - I'll use my brain to do that.
I'm incensed that they're being coy about the fact that Pangram gets paid per words scanned. That's the first reason everyone was livid that they were being forced to scan text to opt out.
I'm also angry about the fact that their scans are so faulty. Make grammatical mistakes or typos? 💯 human. Fix those and suddenly your text is 99% AI.
It's a false and misleading metric.
And if, as you pointed out at the start, it was already easy to detect that something was badly written AI...why do we need a machine to tell us something that we should be learning to figure out for ourselves?
Detecting human words from machine language is a skill we should be honing, not one we should be outsourcing to...machines.
Moreover... if the ideas are human inspired and the content interesting and relevant, who cares if an AI put the words together?
Ai isn't inherently evil. It's a tool. People who use it to spit out regurgitated content scraped from other sources or to make a quick buck off unsuspecting consumers are evil...Pangram isn't going to help us weed those out. It's just going to keep good writers from risking their reputations on a very fallible tool with shady business practices.
I know others hgave stated this, but I also test this to see how accurate this tool is.
I created an article in Claude. 100% AI. Ran the /humanizer skill on the article, then copied and pasted it AS IS into Substack
Ran the AI detector
100% human
It is 100% AI
This is the article: https://proudboomerwellness.substack.com/p/hitting-220-grams-of-protein-under
The issue is it is not correct, and people being people, they are going to start trolling and calling people out when it might not be true. This is a slippery slope.
I have entered my "How I made this ... " in full snark mode, but people screaming about writers using AI by using an AI tool is just ... dumb ... in my opinion
As I've said elsewhere, my main concern is that it could falsely accuse writers of using AI, as these tools aren't reliable and often have false positives. However, one good thing is that people will probably be much less likely to use AI (and there will be a lot less slop).
Unless you run Humanizer inside Claude on it, as I did to test this out ...
HI Kristina. I’ve spouted off enough about this in the past 48 hours but your thoughtful post and the inputs from Patrick and others are helpful color.
I’ve no issue with Substack making this a feature. If an Author/Publisher FEELS they want to have this capability on their publications that’s fine. But it should not be a global feature. Making it a manual opt out EVER atone is a degree of friction that is either designed in deliberately OR the thoughtless product management.
They need to rectify this. and put it back in the hands of the creators decision.
On the contractual position and lazy SLA language and process to notify users this has become extremely prevalent in the new generation of companies. Old norms of what a customer relationship meant and looked like seem to be being deprecated daily. Sad 😞
Good questions, disturbing news about the failure to inform members of the contractual changes. I'm following with interest. I won't use Pangram to check other people's work - I'll use my brain to do that.
I'm incensed that they're being coy about the fact that Pangram gets paid per words scanned. That's the first reason everyone was livid that they were being forced to scan text to opt out.
I'm also angry about the fact that their scans are so faulty. Make grammatical mistakes or typos? 💯 human. Fix those and suddenly your text is 99% AI.
It's a false and misleading metric.
And if, as you pointed out at the start, it was already easy to detect that something was badly written AI...why do we need a machine to tell us something that we should be learning to figure out for ourselves?
Detecting human words from machine language is a skill we should be honing, not one we should be outsourcing to...machines.
Moreover... if the ideas are human inspired and the content interesting and relevant, who cares if an AI put the words together?
Ai isn't inherently evil. It's a tool. People who use it to spit out regurgitated content scraped from other sources or to make a quick buck off unsuspecting consumers are evil...Pangram isn't going to help us weed those out. It's just going to keep good writers from risking their reputations on a very fallible tool with shady business practices.
Sorry. I'm having a lot of feels about this.